DESCRIPTION (Taken from application abstract): Precise therapy with potentially toxic drugs demands precise dosage regimens to achieve and maintain selected therapeutic goals precisely, minimizing variability in patient response. We developed methods to model population drug behavior, determining the entire parameter joint probability density, capable of discovering subgroups of fast and slow metabolizers, for example, when not anticipated. This NPEM program provides useful databases of drug behavior. We coupled it with our Multiple Model (MM) dosage designer for drug dosage decision support. It specifically minimizes variability in patient response about the selected goal(s). We have now developed similar software for implementing on supercomputers large linear and nonlinear pharmacokinetic and pharmacodynamic (PK/PD) population models of multiple drugs and their effects. We will build on these successes by implementing the MM control method for these large models, and an actively adaptive MM controller for smaller linear models. The active controller rehearses future scenarios in advance. It uses the dose as well as serum levels to learn the patient's model to best achieve desired goal(s). It optimizes learning about the patient while treating him at the same time. We will also develop stochastic strategies for optimal experimental design and therapeutic drug monitoring (TDM). One then can optimize coordinated PK/PD therapy with combination drug regimens for cancer, AIDS, bacterial, and viral infections. We will develop software for simulation of clinical drug trials to optimize their design, including concentration-controlled and effect-controlled trials. The MM control and optimal experimental design features will further optimize clinical trial design, exposing fewer patients to risk and reducing drug development costs. We will implement discrete MM parameter distributions from literature data of means, standard deviations, and ranges, for use with the MM controllers, and enhance our clinical MM program with linked general models of effect, diffusion into endocardial vegetations, and bacterial (and viral) growth and kill. Lastly, we will use telemedicine techniques to make both the clinical MM PC programs and the larger clinical PK/PD supercomputer programs available for collaborative telecomputing and videoconferencing. Both caller and consultant will see the same computer output and also each other. This will enhance and disseminate consulting, training, and optimal drug therapy in areas remote from the PC or supercomputer.